The Effects of Detailing on Prescribing Decisions under Quality Uncertainty - PowerPoint PPT Presentation

Loading...

PPT – The Effects of Detailing on Prescribing Decisions under Quality Uncertainty PowerPoint presentation | free to view - id: 17bc86-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

The Effects of Detailing on Prescribing Decisions under Quality Uncertainty

Description:

How does the effectiveness of detailing change when additional information for ... for product level panel data on sales volume, prices, and detailing efforts. ... – PowerPoint PPT presentation

Number of Views:37
Avg rating:3.0/5.0
Slides: 40
Provided by: andrew230
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: The Effects of Detailing on Prescribing Decisions under Quality Uncertainty


1
The Effects of Detailing on Prescribing Decisions
under Quality Uncertainty
  • Andrew Ching
  • Masakazu Ishihara
  • Rotman School of Management
  • University of Toronto

2
Why are there uncertainty about drug quality?
  • Many serious Adverse Drug Reactions (ADRs) are
    discovered only after a drug has been on the
    market for years. Only half of newly discovered
    serious ADRs are detected and documented in the
    Physicians Desk Reference within 7 years after
    drug approval.
  • Lasser et al. (2002) Journal of American Medical
    Association.

3
Pharmaceutical Detailing
  • Detailing sales reps from drug manufacturers
    visit doctors to discuss compliance information,
    side-effects, and efficacy studies.
  • In 2003, detailing costs 8 billion dollars
    journal advertising costs 0.46 billion dollars
    direct-to-consumer (DTC) advertising costs 3.2
    billion dollars.
  • Observations in the pharmaceutical industry
    related to our paper.
  • Uncertainty about drugs qualities
  • Large amount of information about drugs for
    physicians

4
  • Observation
  • Slow diffusion of new drugs suggests learning
    is important.

5
Number of active drugs in Cardiovasculars
  • It is hard for physicians to keep track of the
    latest information about all the drugs.
  • Some physicians may be busy and rely on the
    information provided by detailing.

6
Research Questions
  • How does the effectiveness of detailing change
    when additional information for drugs is revealed
    via patients experiences during the drug
    lifecycle?
  • We develop a structural model of detailing and
    pharmaceutical demand that incorporates learning
    and long-lived effect of detailing.
  • Our model does not assume firms know the true
    quality of their drugs.
  • Designed for product level panel data on sales
    volume, prices, and detailing efforts.

7
Model
  • Agents physicians, firms, and a public health
    agency.
  • There are J products.
  • Each firm has one product.
  • There is one outside alternative (0).
  • Two product characteristics price (pj), and
    quality (qj).
  • Let I(t) (I1(t),…,IJ(t)), be the information
    sets for q.
  • Let Ij be the initial prior for qj when drug j is
    first introduced.
  • Firms observe I(t).
  • Physicians are either well-informed about drug j
    (Ij(t)),
  • or uninformed about drug j (Ijc).
  • Let Mjt be the measure of well-informed
    physicians for drug j at time t. Mjt f(Mjt-1 ,
    Dt).

8
Model (contd)
  • Each period has three stages.
  • (i) Firms observe I(t), and choose Djt.
  • (ii) Mjt is determined for all j, and each
    physician makes his/her prescribing decisions to
    maximize the expected utility for each of his/her
    patient.
  • (iii) Patients consume the drugs and their
    experience signals are revealed to the public
    health agency. The public health agency updates
    I(t) in a Bayesian manner.

9
Bayesian updating of the public information set
  • Experience signal qijt qj dijt,
  • where dijt N(0, s2d).
  • Initial prior for qj N(qj, s2).
  • Expected quality
  • EqjI(t1) EqjI(t) ?j(t)(qjt
    EqjI(t)),
  • where qjt is the sample mean of experiences
    signals revealed for product j in period t.
  • Perception variance
  • s2j(t1) 1 / (1/s2j(t) ?njt/s2d),
  • where njt is the quantity sold for drug j in
    time t
  • 0lt?lt1, is a scaling factor.

10
Measure of well-informed physicians
  • Let Mjt be the measure of well-informed
    physicians about drug j at time t. Mjt f(Mjt-1
    , Dt).
  • Let GjtI be the detailing goodwill stock, and FI
    be the depreciation rate.
  • GjtI (1- FI) Gjt-1I Djt.
  • Mjt g(GjtI, G-jtI).
  • E.g., let Rjt ß0 ß1 GjtI,
  • Mjt exp(Rjt) / (1exp(Rjt)).
  • Average rate of forgetting, FM (Mjt f(Mjt ,
    0)) / Mjt., is a non-linear function of M, an
    inverted-U shape.

11
Physician heterogeneity with endogenous weights
  • Suppose that J 2.
  • Four types of physicians who differ in their
    information sets.
  • Measure of physicians with current information
    about both products M1M2 (I1(t), I2(t)).
  • Measure of physicians with current information
    about only one product Mj(1-Mk), for j ? k
    (Ij(t), Ikc).
  • Measure of physicians who do not have current
    information at all (1-M1)(1-M2) (I1c, I2c).
  • Physician heterogeneity evolves endogenously.
  • Allow the model to depart from the IIA
    restriction.

12
Physicians Choice
  • Patient is utility of consuming drug j
  • uijt a1 - exp(-rqijt) - pppjt eijt.
  • If physician h is well-informed about drug j, his
    expected utility of choosing drug j for patient i
    will be
  • EUhijIj(t)
  • a1 - exp(-rEqjIj(t)-1/2r2(s2dsj2(t))) -
    pppjt eijt,
  • If physician h is uninformed about drug j,
  • EUhijIjc
  • a1 - exp(-rqjc-1/2r2(s2d sc2)) - pppjt eijt.

13
Marginal return of detailing
  • Three factors that affect the marginal return of
    detailing
  • Effectiveness of detailing on building the
    measure of well-informed physicians
  • Changes in the choice probability of physicians
    who are switched from uninformed to informed
    depends on I(t)
  • Measure of well-informed physicians for opponent
    drug.

14
e.g., ß0 -1.4, ß1 5.8e-5, FI 0.03
  • Heterogeneous individual rate of forgetting
  • Potential interactions among physicians

15
Identification
  • Let I Ic
  • Initial market shares identify the initial prior
    mean qualities and variances (qj,s).
  • In the long run, fluctuations of market shares
    and cumulative detailing stocks identify the
    detailing stock parameters (ß0, ß1, FI) and true
    mean qualities (q).
  • After controlling for the evolution of measure of
    well-informed physicians, the evolution of market
    shares over time identifies other learning
    parameters (r, sd).

16
Testable Empirical Implications
  • If initial market share is close to zero
  • Controlling for the cumulative detailing of
    drugs, the marginal return of detailing for drug
    j is positively correlated with own market share
    and negatively correlated with market share of
    opponent drug.
  • When opponent drug keeps improving, the marginal
    return of detailing for own drug is negatively
    correlated with the cumulative detailing of
    opponent drug.

17
Simultaneity Problem
  • We assume that firms observe I(t) before
    detailing takes place in each period. Therefore,
    EqjI(t) may be correlated with Djt.
  • For instance, if EqjI(t) is high, the firm may
    want to assign more detailing efforts to the drug
    j to disseminate the information.
  • Ignoring this correlation will lead to upward
    bias of the parameters associated with detailing.

18
Estimation Strategy
  • Standard estimation strategy is to use BLPs GMM
    approach -- however, hard to use in practice.
  • We follow Ching (2000) (2004) approach.
  • Let sjt (EqjI(t), sj(t), Mjt-1).
  • Djt dj(sjt, s-jt)?jt, where ?jt is the
    prediction error.
  • log(Djt) log(dj(sjt, s-jt)) log(?jt).
  • Use some flexible functional form to approximate
    log(dj(.)).
  • Jointly estimate this pseudo-detailing policy
    function with the demand model.
  • Need to integrate out the unobserved state
    variables simulated maximum likelihood.

19
Actual functional form used
  • We consider J 2.
  • log Djt ?j0(?j1?j2M-jt-1)(1-Mjt-1)?ujtq
    I(?ujtqgt0)
  • (?j3?j4M-jt-1)Mjt-1?ujtq I(?ujtqlt0)
  • ?j5IVjt ?jt.
  • where
  • ?ujtq EujtqI(t) - Eu-jtqI(t)
  • EujtqI(t) - exp(-rEqjI(t)-1/2r2(s2dsj2(
    t)))
  • IVjt total detailing minutes at t by firm j
    in the
  • cardiovascular drug category net
    Djt
  • The higher the expected quality difference, the
    more a firm may want to detail.
  • The smaller the measure of well-informed
    physicians, the higher the incentive to detail.

20
Data
  • Monthly Canadian data on detailing, revenue and
    number of prescriptions from March 93 to Feb 99
    for ACE-inhibitor with diuretic from IMS Canada.
  • Why Canada?
  • Subject to price regulation Patented Medicine
    Prices Review Board.
  • Why ACE-inhibitor with diuretic?
  • No Direct-to-consumer advertising.
  • Only two dominant drugs (Vaseretic and
    Zestoretic).
  • Treat high-blood pressure patients/physicians
    are likely to be risk averse.
  • Market size ACE-inhibitors, ACE-inhibitors w/
    diuretic, and Diuretics, Thiazide.

21
Estimates for Learning, Preference, and Detailing
Stock Parameters
1 Vaseretic (incumbent) 2 Zestoretic (entrant)
22
Parameter Estimates for Pseudo-Detailing Policy
Functions
23
(No Transcript)
24
D 1300, FI 0
25
  • Next, lets look at the goodness-of-fit.

26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
How does the effectiveness of detailing vary over
time?
  • We simulate the effects of a one-time increase in
    detailing for three scenarios.
  • (i) t1, when the expected quality of vaseretic
    (incumbent) is higher
  • (ii) t23, when the expected quality are about
    the same for both drugs
  • (iii) t60, when the expected quality of
    zestoretic (new entrant) is higher.
  • Set Gj0I 24500,which translates to Mj0 0.5.
  • Set Djt 1300, for j1,2, and tgt0.

32
Effectiveness of Detailing Effect of a one-time
increase in detailing by 50 on current demand
33
Conclusion
  • Our model is able to generate a flexible
    diffusion pattern.
  • We quantify the return of detailing at different
    points in time and show it depends on the measure
    of well-informed physicians and the information
    sets.
  • We find evidence that the endogeneity problem
    biases the estimates of the coefficients
    associated with detailing.
  • We find evicence that the role of
    detailing-in-utility becomes much smaller after
    modeling detailing as a means to build/maintain
    the measure of well-informed physicians. For
    this bulletin point, do you plan to talk about
    this during the presentation or you simply
    mention it at the end.

34
Summary Statistics
35
(No Transcript)
36
(No Transcript)
37
  • Observation
  • Detailing expenditures are higher than revenues
    during the
  • Initial stage of the drug lifecycle suggests
    detailing may
  • have long-lived effect.

38
(No Transcript)
39
(No Transcript)
About PowerShow.com